Congestion control with multiple congestion control policy for large bandwidth‐delay‐product networks

Abstract

Recently, broadband applications such as data grids and IP-Storage Area Network have been proliferating. In these applications it is necessary to exchange large amounts of data quickly over networks with a capacity of several gigabits per second. For this purpose, various kinds of congestion control have been proposed for use in such broadband networks. In these kinds of congestion control, the idle bandwidth of the network can be used quickly by increasing the communications speed at the time of communications. On the other hand, complicated forms of congestion are generated in the network. Therefore, when these kinds of congestion control and the existing TCP communications coexist, the existing TCP communication is strongly compressed. In the present paper, TCP with multiple congestion control policies (TCP-MP) is proposed in order to make swift use of idle bandwidths in the network. In TCP-MP, an additional congestion control policy is provided in addition to using a form of congestion control similar to that in conventional TCP, so that information on the variation of the RTT between the transmitter and the receiver can be used to estimate the idle bandwidth of the network and the communications speed is chosen in such a way that the idle bandwidth is used up. Hence, it is possible to quickly use the bandwidth of the network within the range in which the existing TCP communications is compressed. © 2006 Wiley Periodicals, Inc. Electron Comm Jpn Pt 1, 90(1): 45–57, 2007; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecja.20310